The SharePoint nobody could search

That is the naming convention I worked out for Gauzy's SharePoint, running. Paste in filenames the way they actually arrive and it will rewrite them and say which library they belong in. It flags the ones a person still has to look at, because a tool that guesses silently is worse than no tool.

There were about nine thousand files in there and no agreed way to name any of them. The same contract existed as Contract_FINAL.docx, Contract_FINAL_v2.docx, and Contract_FINAL_v2 (1).docx. Nobody could tell which was signed. People stopped searching and started asking me instead, which is how I became the search bar.

I did not want to rename nine thousand files by hand, so I worked out the rules first. Every filename really only needs four things: when it is from, what kind of document it is, what it is about, and which version. Date first, so a folder sorts itself. Type second, so the library it belongs in is obvious. The word FINAL carries no information once there is a version number, so it goes.

Then I used AI for the boring half. I gave it a few hundred of the worst filenames and asked it to find every date format in use. It came back with eight, including 03052024 and 10.25.24, which I would not have thought to check for. I turned those into rules, ran the batch, and reviewed only the files the rules could not place. That last pile was about two hundred, not nine thousand.